We propose a normalized least mean squares algorithm with variable step size. Unlike other solutions, it has low computational cost, only three parameters that are simple to choose, and its ...